DeepMind's Hassabis calls for a U.S.-led global AI watchdog

Demis Hassabis, CEO of Google DeepMind, used a high-profile interview to call for a U.S.-led global AI watchdog, arguing that humanity has only a 'precious window' to establish AGI safety standards before capabilities outpace governance. It's a notable position from a frontier-lab leader, effectively asking for an international coordinating body akin to what exists for nuclear or aviation safety.

Hassabis grounded the appeal in concrete recent chaos rather than abstractions. He cited export-control episodes that whipsawed the industry: Anthropic's most powerful models being frozen 'overnight,' and OpenAI restricting GPT-5.6 to government-vetted partners before its public release. His argument is that ad hoc, unilateral government interventions create instability that a coordinated standards body could smooth.

Competitively and politically, this is a savvy move. By advocating a U.S.-led structure, Hassabis aligns with Washington's desire to lead AI governance while positioning Google as a responsible incumbent — a contrast to the 'move fast' framing associated with some rivals. It also comes as the Trump administration reportedly pressures Anthropic over its Fable model and as distillation disputes multiply.

The skeptical view: calls for regulation from dominant players can double as moat-building, raising compliance costs that favor incumbents over open-source challengers — a tension Hugging Face's Clem Delangue pointedly flagged this week, arguing power shouldn't be 'controlled by just 4 men.' Whether any government acts on Hassabis's proposal, versus treating it as PR, is the open question.